victoriametrics.com
VictoriaMetrics
Anomaly Detection
2024-12-12
Simple, Reliable, Efficient Monitoring
Presented by Fred Navruzov
Agenda
victoriametrics.com
Simple, Reliable, Efficient Monitoring
What Q3’24 ended on
What Q4’24 brought to the table
Plans for Q1’25
What Q3’24 ended on
victoriametrics.com
Simple, Reliable, Efficient Monitoring
Releases (15): 1.13.3 - 1.16.1
Reader:
Allow processing larger queries that exceed
-search.maxPointsPerTimeseries constraint in VictoriaMetrics
Add per-query parameters, like data_range or step to override the global parameters
Improve performance on multicore instances for reading and data processing
Introduced data disk dumps - see how to enable it here
Models:
Add classes supporting online learning on stream-like data, see our documentation
Online z-score, Online MAD, Online Quantile
Added a groupby argument for logical grouping in multivariate models.
What Q4’24 brought to the table
victoriametrics.com
Simple, Reliable, Efficient Monitoring
Releases (14): 1.16.2 - 1.18.8
Reader/Writer/Scheduler/Monitoring:
Added support for multitenant queries, enabling detecting anomalies across
multiple tenants - see our docs for details
Added mutual TLS (mTLS) support in vmanomaly components
Added max_points_per_query arg in VmReader to prevent search.maxQueryDuration
timeouts on longer fit windows and heavier queries
VmReader is now timezone aware (both the reader and query levels) making local
seasonality more accurate, especially during daylight savings changes - see this FAQ section
PeriodicScheduler can be configured to start at a specific time using the start_from and tz
parameters - see this section
What Q4’24 brought to the table
victoriametrics.com
Simple, Reliable, Efficient Monitoring
Models:
Added timezone-aware seasonality support to Prophet and OnlineSeasonalQuantile
models. Should be used in combination with tz arg in Reader section
Improved fit stage performance on multicore instances
(WIP) Improving infer performance on multicore instances stage
System:
Now service can be run on a config directory that is recursively merged and validated
Added config splitting util to divide parent config into fully operational sub-configs -
see this section for the details
What Q4’24 brought to the table
victoriametrics.com
Simple, Reliable, Efficient Monitoring
Self-monitoring to monitor vmanomaly service health
Accompanied by alerting rules
Grafana dashboard for monitoring
Docs & quickstart guide
Based on reworked self-monitoring metrics
Documentation updates
Command-line arguments section in QuickStart
Config splitting section and timezone handling section in FAQ
Monitoring page - now has both logs & metrics for clarity and easier navigation
Plans for Q1’25
Convenience of Use
VMUI-based GUI for testing various anomaly detection configurations on historical data
(WIP) Node-Exporter preset v2.0 with improved visuals and resource-effective online models
(WIP) “How-to” Guides to reduce time to value on vmanomaly
Simple, Reliable, Efficient Monitoring victoriametrics.com
Plans for Q1’25
Resource Efficiency & Scalability
(WIP) Benchmarking utility to precisely estimate existing configurations
Semi-automated horizontal scalability & support in VM Operator
Simple, Reliable, Efficient Monitoring victoriametrics.com
(WIP) Capacity calculator for resource (CPU, RAM, speed) estimation
Simple, Reliable, Efficient Monitoring victoriametrics.com
Plans for Q2’25+
New Frontiers
Integration with VictoriaLogs - anomaly detection on log-based metrics and logs data
Simple, Reliable, Efficient Monitoring victoriametrics.com
Simple, Reliable, Efficient Monitoring victoriametrics.com
Get Started Here
victoriametrics.com/products/enterprise/anomaly-detection/
Product Page
docs.victoriametrics.com/anomaly-detection/
Documentation

VictoriaMetrics Anomaly Detection Updates: Q4 2024

  • 1.
  • 2.
    Agenda victoriametrics.com Simple, Reliable, EfficientMonitoring What Q3’24 ended on What Q4’24 brought to the table Plans for Q1’25
  • 3.
    What Q3’24 endedon victoriametrics.com Simple, Reliable, Efficient Monitoring Releases (15): 1.13.3 - 1.16.1 Reader: Allow processing larger queries that exceed -search.maxPointsPerTimeseries constraint in VictoriaMetrics Add per-query parameters, like data_range or step to override the global parameters Improve performance on multicore instances for reading and data processing Introduced data disk dumps - see how to enable it here Models: Add classes supporting online learning on stream-like data, see our documentation Online z-score, Online MAD, Online Quantile Added a groupby argument for logical grouping in multivariate models.
  • 4.
    What Q4’24 broughtto the table victoriametrics.com Simple, Reliable, Efficient Monitoring Releases (14): 1.16.2 - 1.18.8 Reader/Writer/Scheduler/Monitoring: Added support for multitenant queries, enabling detecting anomalies across multiple tenants - see our docs for details Added mutual TLS (mTLS) support in vmanomaly components Added max_points_per_query arg in VmReader to prevent search.maxQueryDuration timeouts on longer fit windows and heavier queries VmReader is now timezone aware (both the reader and query levels) making local seasonality more accurate, especially during daylight savings changes - see this FAQ section PeriodicScheduler can be configured to start at a specific time using the start_from and tz parameters - see this section
  • 5.
    What Q4’24 broughtto the table victoriametrics.com Simple, Reliable, Efficient Monitoring Models: Added timezone-aware seasonality support to Prophet and OnlineSeasonalQuantile models. Should be used in combination with tz arg in Reader section Improved fit stage performance on multicore instances (WIP) Improving infer performance on multicore instances stage System: Now service can be run on a config directory that is recursively merged and validated Added config splitting util to divide parent config into fully operational sub-configs - see this section for the details
  • 6.
    What Q4’24 broughtto the table victoriametrics.com Simple, Reliable, Efficient Monitoring Self-monitoring to monitor vmanomaly service health Accompanied by alerting rules Grafana dashboard for monitoring Docs & quickstart guide Based on reworked self-monitoring metrics Documentation updates Command-line arguments section in QuickStart Config splitting section and timezone handling section in FAQ Monitoring page - now has both logs & metrics for clarity and easier navigation
  • 7.
    Plans for Q1’25 Convenienceof Use VMUI-based GUI for testing various anomaly detection configurations on historical data (WIP) Node-Exporter preset v2.0 with improved visuals and resource-effective online models (WIP) “How-to” Guides to reduce time to value on vmanomaly Simple, Reliable, Efficient Monitoring victoriametrics.com
  • 8.
    Plans for Q1’25 ResourceEfficiency & Scalability (WIP) Benchmarking utility to precisely estimate existing configurations Semi-automated horizontal scalability & support in VM Operator Simple, Reliable, Efficient Monitoring victoriametrics.com (WIP) Capacity calculator for resource (CPU, RAM, speed) estimation
  • 9.
    Simple, Reliable, EfficientMonitoring victoriametrics.com
  • 10.
    Plans for Q2’25+ NewFrontiers Integration with VictoriaLogs - anomaly detection on log-based metrics and logs data Simple, Reliable, Efficient Monitoring victoriametrics.com
  • 11.
    Simple, Reliable, EfficientMonitoring victoriametrics.com Get Started Here victoriametrics.com/products/enterprise/anomaly-detection/ Product Page docs.victoriametrics.com/anomaly-detection/ Documentation